Amakilabhu amahlaya yizindawo engizikhonzile ngempelasonto. Hlanganyela abangani, ubambe iziphuzo ezimbalwa, futhi uma ihlaya lihlala ngathi sonke—kunesikhathi somlingo lapho amehlo ethu ehlangana khona, futhi sihlanganyela ngokusineka okunezihlathi.

Ukumamatheka kungenza abantu ongabazi babe abangane abathandekayo kakhulu. Iyakhuthaza hlangana-muhle I-Hollywood iceba, ilungisa ubuhlobo obuphukile, futhi ixhunywe ngokungenakuhlukaniswa nemizwa efudumele, efudumele yenjabulo.

Okungenani kubantu. Kumarobhothi, imizamo yawo yokumamatheka kwangempela ivame ukuwela esigodini esiyinqaba—esisondele ngokwanele ukuba sifane nomuntu, kodwa sibangele ukuthintwa. Ngokunengqondo, uyazi ukuthi yini abazama ukuyenza. Kodwa imizwa yamathumbu ikutshela ukuthi kukhona okungalungile.

Kungase kube ngenxa yesikhathi. Amarobhothi aqeqeshelwe ukulingisa isimo sobuso sokumamatheka. Kodwa abazi ukuthi bazoyivula nini i-grin. Lapho abantu bexhuma, siyamamatheka ngokweqiniso ngokuhambisana ngaphandle kokuhlela okuqaphelayo. Amarobhothi athatha isikhathi ukuhlaziya isimo sobuso bomuntu ukuze aphinde aveze ukusineka. Kumuntu, ngisho nama-milliseconds okulibaziseka aphakamisa izinwele ngemuva kwentamo—njengefilimu ethusayo, okuthile kuzwakala kukhohlisayo futhi kungalungile.

Ngesonto eledlule, ithimba e-Columbia University wabonisa i-algorithm lokho kufundisa amarobhothi ukwabelana ngokumomotheka nabasebenzi bawo abangabantu. I-AI ihlaziya izinguquko ezincane zobuso ukuze ibikezele amagama o-opharetha ayo mayelana nama-millisecond angu-800 ngaphambi kokuba zenzeke—isikhathi esanele sokuthi irobhothi libuye lisineke.

Ithimba liqeqeshe ubuso obuthambile be-robotic humanoid obubizwa nge-Emo ukuze lilindele futhi lifanise izinkulumo zomngane walo ongumuntu. Ngobuso be-silicone obunemibala eluhlaza okwesibhakabhaka, i-Emo ibukeka njengenganekwane yesayensi yama-60s. Kodwa yavele yahleka kanye nozakwabo ongumuntu kubude obufanayo “bomzwelo” wamaza.

Amarobhothi e-humanoid avame ukufiphala futhi amise lapho exhumana nabantu, wabhala UDkt. Rachael Jack waseNyuvesi yaseGlasgow, owayengahileleki kulolu cwaningo. I-ChatGPT namanye ama-algorithms olimi amakhulu angenza inkulumo ye-AI izwakale njengomuntu, kodwa ukuxhumana okungenazwi kunzima ukuphindaphinda.

Ukuhlela amakhono okuxhumana nabantu — okungenani okobuso bobuso — kumarobhothi aphathekayo kuyisinyathelo sokuqala sokusiza “amarobhothi omphakathi ukuthi ajoyine umhlaba wezenhlalo yabantu,” wabhala.

Ngaphansi kweHood

Kusukela i-robotaxis kumaseva e-robo akulethela ukudla neziphuzo, amarobhothi azimele ziya ngokuya zingena ezimpilweni zethu.

ELondon, eNew York, eMunich, naseSeoul, ezizimele amarobhothi ngena ezikhumulweni zezindiza ezinesiphithiphithi ezinikeza usizo lwamakhasimende—ukungena, ukuthola isango, noma ukubuyisela imithwalo elahlekile. E-Singapore, amarobhothi amaningana angamafidi ayisikhombisa ubude anombono we-360-degree zulazula esikhumulweni sezindiza ukumaka izinkinga zokuphepha ezingaba khona. Ngesikhathi sobhubhane, izinja zamarobhothi ukuqhelelana okuphoqiwe.

Kodwa amarobhothi angenza okwengeziwe. Emisebenzini eyingozi—njengokuhlanza imfucumfucu yezindlu ezicekelwe phansi noma amabhuloho—bangaphayona emizamweni yokuhlenga futhi bandise ukuphepha kwabafika kuqala. Ngenxa yokukhula kwenani labantu emhlabeni wonke, bangasiza abahlengikazi ukuba basekele asebekhulile.

yamanje amarobhothi humanoid zithandeka ngendlela ekhathuni. Kodwa isithako esiyinhloko sokuthi amarobhothi angene emhlabeni wethu ukwethembana. Njengoba ososayensi bakha amarobhothi anobuso obufana nobomuntu oba buyanda, sifuna ukusho kwabo kufane nesikulindele. Akukhona nje ngokulingisa isimo sobuso. Ukumamatheka okwabiwe kwangempela kokuthi “yebo ngiyazi” phezu kwehlaya elifanele ukuxega kwakha isibopho.

Ukuxhumana okungakhulumi—izinkulumo, ukuthinta kwezandla, ukuma komzimba—amathuluzi esiwasebenzisayo ukuze siziveze. Nge-ChatGPT nezinye i-AI ekhiqizayo, imishini “ingakwazi kakade ukuxhumana ngevidiyo nangamazwi,” wathi cwaningo umbhali uDkt. Hod Lipson ukuze Isayensi.

Kodwa uma kuziwa emhlabeni wangempela—lapho ukubuka, ukucwayiza, nokumamatheka kungenza umehluko—“kuyisiteshi esingekho njengamanje,” kusho uLipson. “Ukumamatheka ngesikhathi esingalungile kungase kubangele umonakalo. [Uma ngisho nama-millisecond ambalwa sekwephuze kakhulu], kuzwakala sengathi uyazulazula mhlawumbe.”

Kusho uCheese

Ukuze benze amarobhothi enze isenzo esingakhulumi, ithimba ligxile esicini esisodwa—ukumomotheka okwabiwe. Izifundo zangaphambilini zinamarobhothi ahlelwe ngaphambilini ukuze alingise ukumamatheka. Kodwa ngenxa yokuthi azizenzeki, kubangela ukubambezeleka okuncane kodwa okuqaphelekayo futhi kwenza i-grin ibukeke ingelona iqiniso.

“Ziningi izinto ezingena ekuxhumaneni okungakhulunywanga ngomlomo” okunzima ukuzibala, kusho uLipson. "Isizathu sokuthi sidinga ukuthi 'shizi' lapho sithatha isithombe yingoba ukumamatheka lapho sifunwa khona empeleni kunzima kakhulu."

Ucwaningo olusha lugxile esikhathini.

Ithimba lenze i-algorithm elindele ukumamatheka komuntu futhi lenze ubuso obunjengobomuntu be-animatronic busineke buhambisana. Ubuso berobhothi obubizwa ngokuthi i-Emo, bunamagiya angu-26—cabanga imisipha yokwenziwa—embozwe “isikhumba” esinwebekayo se-silicone. Igiya ngalinye lixhunywe “kuhlaka lwamathambo” oluyirobhothi olunozibuthe ukuze lunyakazise amashiya, amehlo, umlomo nentamo. Amehlo ka-Emo anamakhamera akhelwe ngaphakathi ukuze aqophe indawo ezungezile futhi alawule ukunyakaza kwenhlamvu yeso kanye nokucwayiza.

Ngokwakhe, i-Emo ingakwazi ukulandelela isimo sayo sobuso. Umgomo walolu cwaningo olusha wawuwukusiza ukuhumusha imizwelo yabanye. Ithimba lisebenzise iqhinga noma yimuphi umuntu osemusha osemusha okungenzeka uyazi: Bacele u-Emo ukuthi abheke esibukweni ukuze afunde ukulawula amagiya akhe futhi akhe isimo sobuso esiphelele, njengokumomotheka. Irobhothi lafunda kancane kancane ukufanisa indlela elisho ngayo nemiyalelo yezimoto—ake uthi, “phakamisa izihlathi.” Ithimba libe selisusa noma iluphi uhlelo olungase lunwebe kakhulu ubuso, kulimaze isikhumba se-silicon serobhothi.

“Kuvele ukuthi…[ukwenza] ubuso berobhothi obukwazi ukumamatheka bekuyinselele emangalisayo uma ubuka izinto. Kunzima kunokwenza isandla serobhothi,” kusho uLipson. “Sisebenza kahle kakhulu ekuboneni ukumamatheka okungelona iqiniso. Ngakho siyazwela kakhulu kulokho.”

Ukuze kulwisane nalesi sigodi esingaqondakali, ithimba laqeqesha u-Emo ukuba abikezele ukunyakaza kobuso kusetshenziswa amavidiyo abantu abahlekayo, bemangele, beshwabene, bekhala, futhi bekhuluma ezinye izimo. Imizwa ikhona yonke indawo: Uma umomotheka, amakhona omlomo wakho agobeka abe yinyanga ecwebile. Uma ukhala, izinzipho zilandelana.

I-AI ihlaziye ukunyakaza kobuso besithombe sohlaka ngalunye ngohlaka. Ngokulinganisa ibanga phakathi kwamehlo, umlomo, nezinye “izimpawu zomhlaba zobuso,” ithole izimpawu ezihambisana nomzwelo othile—ngokwesibonelo, ukuphakama kwekhona lomlomo wakho kusikisela ukumomotheka, kuyilapho ukunyakaza okuya phansi kungase kubonise ukumomotheka. ehle ahwaqabale.

Uma isiqeqeshiwe, i-AI yathatha isikhathi esingaphansi kwesekhondi ukubona lezi zimpawu zomhlaba zobuso. Lapho inika amandla i-Emo, ubuso berobhothi bebungalindela ukumomotheka okusekelwe ekusebenzelaneni kwabantu phakathi nesekhondi, kangangokuthi lasineka nomhlanganyeli walo.

Ukuze kucace, i-AI “ayizwa”. Kunalokho, iziphatha ngendlela umuntu angenza ngayo lapho ehleka ukuze asukume ahlekisayo ngokumamatheka okungathi kwangempela.

Isimo sobuso akuzona kuphela izimpawu esizibonayo lapho sixhumana nabantu. Ukunikina ikhanda okucashile, ukunqekuzisa ikhanda, amashiya aphakanyisiwe, noma ukuthinta kwezandla konke kwenza uphawu. Kungakhathaleki amasiko, ama-“ums,” “ahhs,” kanye “nokuthandwa”—noma okufana nawo—kuhlanganiswe ekusebenzelaneni kwansuku zonke. Okwamanje, u-Emo usefana nengane efunde ukumamatheka. Ayikaqondi ezinye izingqikithi.

“Kusekuningi okusazohamba,” wathi Lipson. Sifunda nje ukuxhumana okungakhulumi kwe-AI. Kodwa "uma ucabanga ukuthi ukuzibandakanya ne-ChatGPT kuyathakazelisa, vele ulinde kuze kube yilapho lezi zinto ziba ngokomzimba, futhi konke ukubheja kuvaliwe."

Isikweletu sezithombe: U-Yuhang Hu, Columbia Engineering nge-YouTube

• I-SEO Powered Content & PR Distribution. Khuliswa Namuhla.
• I-PlatoData.Network Vertical Generative Ai. Zinike Amandla. Finyelela Lapha.
• I-PlatoAiStream. I-Web3 Intelligence. Ulwazi Lukhulisiwe. Finyelela Lapha.
• I-PlatoESG. Ikhabhoni, I-CleanTech, Amandla, Environment, Ilanga, Ukuphathwa Kwemfucuza. Finyelela Lapha.
• I-PlatoHealth. I-Biotech kanye ne-Clinical Trials Intelligence. Finyelela Lapha.
• Source: https://singularityhub.com/2024/04/01/this-robot-predicts-when-youll-smile-then-grins-back-right-on-cue/